2015 - Round 2 : Uncertainty and Caution

No sooner, it seems, at this time of year, does one round finish before the next one starts.

It didn't help that the TAB saw fit to have Head-to-Head markets up just a day or so after the final game in Round 1, and Line markets up not all that long after, so I found myself locking in MoS Wagers, Tips and Predictions in a QANTAS lounge in Townsville about a day sooner than I'd expected.

Not that I'm altogether unhappy with what Investors are faced with, which is summarised in the table below.

We've just two Head-to-Head wagers this week, one on an underdog Collingwood and the other on favourites Port Adelaide, coupled with six Line bets, five of them on favourites (including Port Adelaide) offering between 2.5 and 32.5 points start, and one more on the Pies, this week enjoying just under a 2-goal start.

Together those eight bets, which is six fewer bets than last week, represent risks and rewards as summarised in the following Ready Reckoner.

Three games, those with Eagles, Suns and Dons home teams, represent no threat at all for Investors, whilst the Pies v Crows, and Power v Swans games tantalise with upsides of 1.5c or more since each carries both a Head-to-Head and a Line bet on the home team. No other game offers an upside in excess of +0.7c. 

Downside is also, as we'd expect, concentrated in the Pies and the Power, each wielding the ability to knock about 1.5c to 2c off the Portfolio price. Four other teams, Richmond, GWS, Geelong and the Kangaroos, are capable of shaving another 0.8c off each.

All told, a great week would add about 6c to the value of the Overall Portfolio, while a tragicomic round would strip about 6.5c from it.

HEAD-TO-HEAD TIPSTERS

Overall, the level of disagreement amongst Head-to-Head Tipsters this week is very high, the combined average Disagreement Index coming in at 28%, which is more than double Round 1's 12% figure.

No single Tipster, in fact, has an Index below 22%, and the maximum of 40% is associated with three Tipsters: Consult The Ladder, Follow The Streak, and Easily Impressed I. As a group, the Win family of Tipsters also stands out on the basis of its contrarianism (more on which later).

At a game level, disagreement is highest for the Collingwood v Adelaide, Gold Coast v St Kilda, Geelong v Fremantle, and Kangaroos v Brisbane Lions contests where even the lesser-fancied Collingwood, Gold Coast, Fremantle and Brisbane Lions enjoy at least 40% support.

In the other contests, the favoured team has at least 70% support, with Hawthorn, West Coast and Richmond each rallying 90% or more of the Head-to-Head Tipsters behind them.

MARGIN PREDICTION

On margin prediction, the algorithms underpinning the Win, ProPred and H2H Predictors have shown markedly dissenting behaviour this week, the Win pair finding themselves the lone supporters of a Carlton win, at the high end of the margin forecasts for Port Adelaide and Geelong wins, and in the heavy minority tipping a Saints victory, albeit it in their case by the narrowest of margins.

The ProPred duo find themselves also amongst the more bullish Port Adelaide and Geelong spruikers, while the H2H quartet rate Port Adelaide's chances very highly and have the Geelong v Fremantle game as near enough a draw.

As well, Combo_NN1 has unleashed margin predictions of between 11 and 12 goals in two games - Suns v Saints, and Roos v Lions - both of which predictions are 6 goals or more larger than any other MoS Margin Predictor. Combo_NN1 has recorded the most extreme home team win margin prediction in 3 more of the 9 contests and, as a consequence, has this week's highest MAD at almost 18 points per game.

C_Marg, for its part, has recorded the three most extreme away team margin predictions and so finds itself with the 4th-highest MAD of 12.5 points per game, just behind Win_7 (14.3) and Win_3 (13.5). Combo_7 has the week's smallest MAD at just 3.5 points per game, narrowly ahead of Bookie_9, who recorded last week's low-MAD.

Measured either by the range of forecast margins or their MAD, the Cats v Dockers, and Roos v Lions games have generated comfortably the highest levels of disagreement this week, though the MADs for five other games also exceed 6 points. Only three teams, Richmond, Port Adelaide and Hawthorn, enjoy unanimous Margin Predictor support.

The generally high levels of disagreement are reflected in the week's overall MAD of 8.8 points per game, which is almost 65% higher than last week's figure.

PROBABILITY PREDICTION

WinPred's, ProPred's and H2H's dissention is reflected also, as you'd expect, in their probability predictions. So extreme was H2H's assessment of Port Adelaide's chances that the 25% cap needed to be invoked for H2H_Adjusted, making its probability prediction about 3% lower than H2H_Unadjusted's. Note that the 25% adjustment is only made for excessively high home team probability assessments, not for excessively low ones, so the relatively low assessments by H2H of the Gold Coast's and the Kangaroos' chances remain unadjusted.

WinPred has the round's high MAD of 14.5% points per game, though it sits only slightly ahead of C_Prob's 14.1%, H2H_Unadjusted's 13.4%, and H2H_Adjusted's 13.0%. ProPred has the week's low MAD of 6.8% points per game.

Three games have MADs roughly in the 15% points or higher range (Suns v Saints, Roos v Lions, and Pies v Crows), while two have MADs of under 6% points (Eagles v Blues, and Dons v Hawks).

Combined, the nine contests have an average MAD of 10.6% points per game, which is almost 50% higher than last week's figure.

In summary then, there's a lot more disagreement amongst the Tipsters and Predictors this week than there was last week, the main sources of which you can glean from the table at right, which records individual and by-group Tipsters' and Predictors' Disagreement Indexes and MADs for Round 1 and Round 2. 

Amongst the Head-to-Head Tipsters you can see that every Tipster except Home Sweet Home has a higher Disagreement Index in Round 2 than in Round 1. Many of them have significantly higher Index values as discussed earlier.

Looking next at the Margin Predictors we see that only three have recorded smaller MADs in Round 2 than in Round 1 (Combo_7, Combo_NN2 and ProPred_3) and that six Predictors have MADs this week that are more than double those of last week (C_Marg, Combo_NN1, Win_3, H2H_Adj_3, H2H_Unadj_3 and H2H_Unadj_10). Also, where last week saw no Predictor record a double-digit MAD, this week eight of the exceeded that mark.

Finally, reviewing the Probability Predictors we find that only ProPred has recorded a smaller MAD in Round 2 than in Round 1. Four Predictors, C_Prob, WinPred, H2H_Adjusted and H2H_Unadjusted, all recorded MADs for Round 2 that were 60 to 100% higher than their MADs for Round 1.

Of itself, the heightened levels of uncertainty suggested by the greater variability in MoS predictions this week probably doesn't mean much and is likely to be driven in part by the rates at which the different underlying algorithms are recalibrating themselves to the new season. A different set of predictors with different rates of adaptation might have produced the same or even less variability this week than we saw last week. Still, given that these Predictors include those that form the basis of the Head-to-Head and Line Funds, I take some comfort from the fact that wagering is more subdued this week than it was last week, and from the fact that wagering caps for both Funds are kept low for a few more weeks yet.